Machine Learning for Predicting Fatigue Life of Mechanical Components

Mechanical components could suffer fatigue life from functioning during their service period. Machine learning (ML) is a promising approach for forecasting the fatigue life of components. This research offers an effective approach for reliably estimating the fatigue lifespan of mechanical components beneath uniaxial loads in high-cycle fatigue conditions. Different ML methods are used to achieve this across a broad spectrum of components, operating circumstances, notch shapes, and fatigue lifetimes. Conventional techniques for this problem have depended heavily on any of the mechanical reaction characteristics, like strain, stress, or power. This research additionally identifies which of these variables is the best indicator. This study shows that all 3 variables can be used well to estimate fatigue lifespan, with stress-based estimates being the most accurate. Gradient boosters and Random Forest outperform all other ML techniques tested. An outstanding result is a considerable enhancement in prediction accuracy obtained by adding novel data acquired using the Basquin formula.